Introduction
Search strategy and selection criteria
Results
Fall risk classification
Assessment tool | Description |
---|---|
Barthel Index [29] | Ordinal scale that ranks subjects from 0 (total dependence) to 100 (total independence) based on 8 self-care and 2 mobility activities of daily living. |
Fried’s Frailty Criteria [30] | Presence of 3 or more of 5 frailty indicators (significant and unintentional weight loss, grip weakness, poor endurance and energy, slow gait speed, low physical activity level). |
Fukuda Test [31] | The person is blindfolded, extends both arms, and marches in place for 50 to 100 steps. Maximum body rotation greater than 30° indicates vestibular deficits. |
Mini Motor Test [32] | 20 item test that assesses abilities in bed (2 items), sitting position (3 items), standing position (9 items), and gait (6 items). |
One Legged Stance Test [33] | Time a person can stand on one leg without upper extremity support and without bracing the suspended leg against the stance leg. Greater than 30 s indicates low fall risk and less than 5 s indicates high fall risk. |
Physical Performance Test [34] | Ability to stand with feet together side-by-side, semi-tandem, and tandem; walk 8 ft; and rise from a chair and return to seated position. |
Physiological Profile Assessment (PPA) [35] | Assessment of vision, peripheral sensation, muscle force, reaction time, and postural sway. Score of 0-1 = mild risk, 1-2 = moderate risk, and >2 = high risk of falling. |
STRATIFY Score [36] | Assessment of 2 month fall history, mental alteration, frequent toileting, visual impairment, psychotropic medication use, and mobility issues. Score of <2 indicates increased fall risk. |
Timed Up and Go (TUG) [37] | Time to stand up from an armchair, walk 3 m, turn, walk back to the chair, and sit down again. Times that exceed 14 s indicate increased fall risk for community dwelling elderly without neurological disorders. |
Tinetti Assessment Tool [38] | Dynamic balance and gait evaluation with 10 balance components and 8 gait components. Overall scores <19 = high fall risk, 19-23 = moderate fall risk, > 23 = low fall risk. Maximum score = 40. |
Retrospective history | Prospective occurrence | Assessment tools | |
---|---|---|---|
Auvinet et al., 2003 [39] | 1 year | - | - |
Bautmans et al., 2011 [40] | 6 months | - | TUG >15 s or Tinetti score ≤24 |
Caby et al., 2011 [41] | 1 year | - | 25 m walking, Mini Motor test, Tinetti test, TUG, Physical Performance Scale, Fukuda test, One Legged Stance test |
Cho and Kamen 1998 [18] | 1 year | - | Self-reported frequent fallers |
Doheny et al., 2011 [42] | 5 years | - | Self-reported fear of falling or presence of cardiovascular risk factors |
Doheny et al., 2012 [43] | 5 years | - | - |
Doi et al., 2013 [44] | - | 1 year (reported weekly) | - |
Ganea et al., 2011 [45] | - | - | Fried’s criteria for frailty |
Unspecified | - | Tinetti test level 3 | |
Gietzelt et al., 2009 [49] | - | - | STRATIFY score (includes 2 month fall history) ≥2 |
5 years | - | - | |
Ishigaki et al., 2011 [52] | - | - | One Legged Stance test (eyes open) ≤15 s and/or TUG ≥11 s |
Kojima et al., 2008 [53] | 1 year | - | - |
Laessoe et al., 2007 [54] | - | 1 year (fall diary with contact at 6 months) | - |
Latt et al., 2009 [55] | 1 year | - | - |
Liu et al., 2008 [56] | Unspecified | - | Falling during gait perturbation assessment, medical history, self-identification as frequent faller |
Liu et al., 2011 [57] | - | - | PPA |
Liu et al., 2011 [58] | 1 year | - | - |
Marschollek et al., 2008 [59] | - | - | TUG > 20 s, STRATIFY score >2, Barthel Index: Mobility score <10 |
Marschollek et al., 2009 [60] | In-hospital history | - | - |
- | 1 year | - | |
Martinez-Ramirez et al., 2011 [63] | - | - | Body mass loss ≥4.5 kg, low energy, low physical activity, weakness, slowness |
Menz et al., 2003 [64] | - | - | Overall fall risk score (low, moderate, high risk) based on vision, peripheral sensation, strength, reaction time, balance tests |
Moe-Nilssen et al., 2005 [65] | 1 year | - | - |
Najafi et al., 2002 [17] | - | - | Fall risk score ≥5 based on balance, gait, visual, cognitive and depressive disorders, history of falls. |
- | - | PPA | |
O’Sullivan et al., 2009 [1] | 1 year | - | - |
Paterson et al., 2011 [69] | - | 1 year (reported monthly) | - |
Redmond et al., 2010 [70] | - | - | PPA |
Schwesig et al., 2012 [71] | - | 1 year (recorded by caregivers) | - |
Senden et al., 2012 [72] | - | - | Tinetti test ≤24 (Low risk 19-24, High risk <19) |
Toebes et al., 2012 [73] | 1 year | - | - |
Weiss et al., 2011 [74] | 1 year | - | - |
Yack and Berger [75] | 1 year | - | Self report of unsteady or unstable walking and/or standing |
Inertial sensors
Sensor location
Assessed activity
Variables
Category | Variable | Sensor location |
---|---|---|
Position and Angle Variables
| AP peak to peak amplitude | LB [52] |
ML peak to peak amplitude | LB [52] | |
V peak to peak amplitude | LB [52] | |
AP and ML postural sway length during stance | ||
Trunk tilt | St [44] | |
Angular Velocity Variables
| Min, mean, max AP | Sha [50] |
Min, mean, max ML | Sha [50] | |
Min, mean, max V | Sha [50] | |
AP peak to peak amplitude | LB [52] | |
ML peak to peak amplitude | Sha [50] | |
V peak to peak amplitude | LB [52] | |
Postural sway velocity during stance | ||
Mean squared modulus ratio for postural sway | ||
AP RMS during stance | LB [51] | |
ML RMS during stance | LB [51] | |
V RMS during stance | LB [51] | |
3D RMS during stance | LB [51] | |
ML variability | UB [73] | |
Linear Acceleration Variables
| Median AP | LB [74] |
SD of AP | ||
Peak AP | UB [75] | |
Peak V | UB [75] | |
AP peak to peak amplitude | LB [52] | |
ML peak to peak amplitude | ||
V peak to peak amplitude | LB [52] | |
AP RMS | ||
ML RMS | ||
V RMS | ||
AP RMS during stance | LB [51] | |
ML RMS during stance | LB [51] | |
V RMS during stance | LB [51] | |
2D RMS (ML and AP) during stance | LB [43] | |
3D RMS | ||
3D RMS during stance | LB [51] | |
Jerk | St [42] | |
Sit to stand AP range | LB [74] | |
Stand to sit AP range | LB [74] | |
Sit to stand Jerk | LB [74] | |
Dissimilarity of AST subcomponents | ||
Dissimilarity of STS subcomponents | ||
Spatial Variables
| Number of steps | |
Step length | ||
Temporal Variables
| Gait Speed | |
Cadence | ||
Step duration | ||
Step duration variability | ||
Stride time | Fo [71] | |
SD of stride time | Fo [71] | |
% GC double support | Sha [50] | |
TUG time | ||
TUG subcomponent time | ||
TUG: number of gait cycles | Sha [50] | |
STS time | LB [66] | |
STS subcomponent times | ||
SD of STS subcomponent times | LB [66] | |
Normalized SD of STS subcomponent times | ||
Sit/stand transition duration | ||
Sit/stand SD of transition duration | St [17] | |
AST time | LB [66] | |
AST subcomponent times | ||
SD of AST subcomponent times | LB [66] | |
Normalized SD of AST subcomponent times | ||
Energy Variables
| Kinetic Energy | LB [49] |
Local wavelet energy | St [45] | |
Summed magnitude area of acceleration | ||
25% quartile frequency | ||
50% quartile frequency | ||
75% quartile frequency | ||
Sway frequency during stance | LB [51] | |
Number of FFT peaks | ||
Dominant FFT peak parameters | ||
1st FFT peak parameters | ||
Ratio of magnitude of even harmonics to odd harmonics | ||
Area under 1st 6 harmonics divided by remaining area | ||
Ratio of 1st 4 harmonics to magnitude of 1st 6 harmonics | ||
ML spectral edge frequency | St [42] | |
Entropy of power spectrum | LB [53] | |
Correlation between left and right arm signals | ||
Maximum V acceleration Lyapunov Exponent | ||
Maximum AV Lyapunov Exponent | ||
Autocorrelation coefficients of acceleration signal | ||
Trunk level forces | St [45] | |
Continuous wavelet transform | LB [63] | |
Discrete wavelet transform | ||
Detrended fluctuation fractal scaling index of acceleration derived stride time | Fo [69] | |
Fractal dimension of acceleration versus AV | St [45] | |
Number of abnormal sit/stand transitions | St [17] |
Classification models of fall risk prediction
Author | Model | Model validation | Accuracy (%) | Specificity (%) | Sensitivity (%) |
---|---|---|---|---|---|
Caby et al., 2011* [41] | Radial basis function neural network, support vector, k-nearest neighbour, and naive Bayesian classifiers | Leave-one-out cross-validation | 75-100 | 40-100 | 93-100 |
Giansanti et al., 2008*† [48] | Multi-layer perceptron neural network | 47:53 split (Train:Test) | 97 | 97 | 98 |
Giansanti et al., 2006*† [46] | Mahalanobis cluster analysis | 47:53 split (Train:Test) | 93.5-94.5 | 93-94 | 93.9-94.9 |
Giansanti et al., 2008*† [47] | Multi-layer perceptron neural network | 47:53 split (Train:Test) | 88-91 | 88-92 | 88-91 |
Gietzelt et al., 2009* [49] | Decision tree | Not specified | 90.5 | 91.0 | 89.4 |
Ganea et al., 2011* [45] | Logistic regression, ROC curve | Not specified | - | 35-88 | 55-92 |
Weiss et al., 2011† [74] | Logistic regression | Not specified | 63.4-87.8 | 50.0-83.3 | 65.2-91.3 |
Liu et al., 2011* [57] | Linear regression, linear discriminant classifier | Leave-one-out cross-validation | 71 | 98.3 | 88.9 |
Marschollek et al., 2011‡ [61] | Logistic regression, decision tree | Stratified ten-times ten-fold cross validation | 78-80 | 82-96 | 58-74 |
Marschollek et al., 2008* [59] | Logistic regression, classifier | Stratified ten-times ten-fold cross validation | 65.5-89.1 | 15.4-60.4 | 78.5-99.0 |
Marschollek et al., 2009† [60] | Decision tree | Not possible due to limited sample size | 90 | 100 | 57.7 |
Schwesig et al., 2012‡ [71] | Binary logistic regression, ROC curve | Not specified | - | 42-61 | 63-100 |
Moe-Nilssen et al., 2005† [65] | Linear regression, ROC curve | Not specified | 80 | 85 | 75 |
Bautmans et al., 2011† [40] | Logistic regression, ROC curve | Not specified | 77 | 78 | 78 |
Greene et al., 2010† [50] | Logistic regression | 80:20 split (Train:Test) | 76.8 | 75.9 | 77.3 |
Doi et al., 2013‡ [44] | Logistic regression, ROC curve | Not specified | - | 84.2 | 68.8 |
Marschollek et al., 2011‡ [62] | Logistic regression, classifier | Stratified ten-times ten-fold cross validation | 70 | 78 | 58 |
Greene et al., 2012† [51] | Support vector machine | Ten-fold cross validation | 71.5 | 68.4 | 65.4 |
Kojima et al., 2008† [53] | Regression, canonical discriminant classifier | Not specified | 62.1 | 68.2 | 61.1 |
Senden et al., 2012* [72] | Linear regression, ROC curve | Not specified | AUC: 0.67-0.85 | - | - |